Method and system for evaluating state of on-load tap changer of transformer, electronic equipment and medium

Abstract

The application provides a kind of state evaluation method, system, electronic equipment and medium of transformer on-load tap changer, belongs to the fault diagnosis field of transformer on-load tap changer device.The method comprises: obtaining the electrical and mechanical performance parameters of transformer on-load tap changer, generating oil chromatogram data feature matrix for oil chromatogram data;For vibration signal, generate vibration signal feature matrix;For driving motor current signal, generate driving motor current signal feature matrix;Fusion three kinds of feature matrix generate multi-dimensional information feature matrix;The correlation between the data is calculated using the attention mechanism and the multi-dimensional information feature matrix is assigned a weight, and it is input into the BIGRU model for state evaluation.The characteristic data of the electrical characteristics and mechanical characteristics of the on-load tap changer are comprehensively selected, and the evaluation model is used for fault diagnosis, which improves the diagnostic accuracy and adaptability.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of fault diagnosis of transformer on-load tap changer devices, and particularly relates to a transformer on-load tap changer state evaluation method and system, an electronic device and a medium. BACKGROUND

[0002] With the rapid development of China's electric power industry, the equipment capacity of the power grid and the scale of the power grid are also becoming larger and larger, which leads to more high economic losses caused by the fault maintenance of power equipment, and the social effects will become more and more serious, which puts forward higher technical requirements for the protection of power supply safety and quality. Therefore, effective protection measures must be taken to improve the reliability of the normal operation of power equipment and the stability of system operation to effectively ensure the quality of power supply.

[0003] As the only movable component in the transformer, the accurate and timely action of the on-load tap changer (OLTC) can not only reduce and avoid large fluctuations in voltage, but also can force the distribution of load flow, tap the device reactive and active power output, and ensure the safe and reliable operation of the power system. Statistics show that OLTC faults account for about 30% of the total transformer faults, and the fault types are basically mechanical faults, such as contact failure, component loosening, spring fatigue, abnormal switching time, etc. Once the OLTC has a mechanical fault, it may cause line tripping loss of transmission power, or even cause the converter transformer to catch fire and burn out, causing huge economic losses and adverse social impact.

[0004] However, the traditional OLTC state detection method mainly relies on electrical performance measurement, including testing the transition resistance impedance and judging the switching time according to the current on-off situation, and the existing electrical measurement method is often helpless for mechanical fault hidden dangers in the OLTC switching process. For mechanical performance, in recent years, a mechanical state vibration detection method has been proposed, which collects vibration signals by arranging vibration sensors on the surface of the equipment, and realizes state detection by extracting the characteristics of vibration signals under different working conditions. However, the current traditional mechanical fault diagnosis method mainly depends on a single feature, and has problems such as low diagnosis accuracy and poor adaptability. SUMMARY

[0005] The purpose of the embodiments of the application is to provide a transformer on-load tap changer state evaluation method, system, electronic device and medium, which can solve at least part of the technical problems existing in the prior art.

[0006] In a first aspect, the embodiments of the application provide a transformer on-load tap changer state evaluation method, comprising:

[0007] Obtaining electrical performance parameters and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatographic data, and the mechanical performance parameters include a vibration signal and a driving motor current signal;

[0008] Generating an oil chromatographic data feature matrix for the oil chromatographic data, wherein the oil chromatographic data feature matrix includes a volume fraction of a gas generated by decomposition of insulation oil of the on-load tap changer in an abnormal situation, a total hydrocarbon value, a temperature of a thermal fault point of the on-load tap changer, a temperature of a discharge fault point, and an acetylene growth rate;

[0009] Extracting vibration signal features of the hydroelectric unit vibration signal by using a multi-scale wave fluctuation spread entropy for the vibration signal, and introducing energy entropy, kurtosis, and spectral kurtosis to generate a vibration signal feature matrix;

[0010] Generating a driving motor current signal feature matrix for the driving motor current signal, wherein the driving motor current signal feature matrix includes a duration of the driving motor current and a sum of absolute values of the driving motor current in a motor starting stage, a motor stable running stage, and a motor stopping running stage;

[0011] Fusing the oil chromatographic data feature matrix, the vibration signal feature matrix, and the driving motor current signal feature matrix to generate a multi-dimensional information feature matrix;

[0012] Calculating correlations between data in the multi-dimensional information feature matrix by using an attention mechanism and assigning weights to the multi-dimensional information feature matrix to obtain data samples with weights, and inputting the data samples with weights into a pre-constructed BIGRU model to realize state evaluation of the on-load tap changer of the transformer.

[0013] Optionally, the gas generated by decomposition of insulation oil of the on-load tap changer in an abnormal situation includes at least hydrogen, methane, ethane, ethylene, and acetylene.

[0014] Optionally, the temperature of the thermal fault point of the on-load tap changer is calculated according to the following formula:

[0015]

[0016] In the formula, T G is the temperature of the thermal fault point, is the volume fraction of ethylene, is the volume fraction of ethane.

[0017] Optionally, the temperature of the discharge fault point is calculated according to the following formula:

[0018]

[0019] In the formula, for the temperature of the discharge fault point, for the hydrogen volume fraction, for the acetylene volume fraction.

[0020] Optionally, the acetylene growth rate is calculated according to the following formula:

[0021]

[0022] In the formula, for the acetylene growth rate, for the change in acetylene volume fraction, for the time interval.

[0023] Optionally, the calculation process of the multiscale fluctuation entropy includes:

[0024] Mapping the coarse-grained time series, and performing linear transformation on the mapping result to obtain Z;

[0025] Calculating the embedding vector according to the embedding dimension and the time delay, and mapping each embedding vector to a fluctuation-based spreading pattern;

[0026] Calculating the probability of each spreading model, and calculating the fluctuation-based spreading entropy based on the calculation method of Shannon entropy;

[0027] Calculating the fluctuation-based spreading entropy under the scale factor to obtain a multiscale fluctuation spreading entropy function.

[0028] Optionally, the energy entropy is calculated according to the following formula:

[0029]

[0030] In the formula, denotes the vibration amplitude, denotes the ratio of the current position energy to the total energy, denotes the time series.

[0031] Optionally, the kurtosis is calculated according to the following formula:

[0032]

[0033] In the formula, N is the data sample capacity, is the i-th data sample, is the sample mean.

[0034] Optionally, the spectral kurtosis is calculated according to the following formula:

[0035] ;

[0036] ;

[0037] In the formula, representing a frequency, representing a power spectral density of a current frequency, representing an average power spectral density.

[0038] Optionally, the correlation between data in the multi-dimensional information feature matrix is calculated by using an attention mechanism, and a weight is assigned to the multi-dimensional information feature matrix to obtain a data sample with a weight, including an additive attention scoring mechanism:

[0039]

[0040] wherein, is an attention scoring mechanism after one full connection operation of the hidden layer, u s , w , b are respectively a randomly initialized time series, an attention weight matrix and a bias term matrix, is i a hidden layer state of the i-th feature group data. j

[0041] Optionally, the correlation between data in the multi-dimensional information feature matrix is calculated by using an attention mechanism, and a weight is assigned to the multi-dimensional information feature matrix to obtain a data sample with a weight, and further including an attention distribution mechanism:

[0042]

[0043] wherein, is i a degree of attention of the i-th feature group data to the target prediction quantity. j

[0044] Optionally, the correlation between data in the multi-dimensional information feature matrix is calculated by using an attention mechanism, and a weight is assigned to the multi-dimensional information feature matrix to obtain a data sample with a weight, and further including an information weighted summation mechanism:

[0045]

[0046] wherein, is a score vector after weighted summation, is i a degree of attention of the i-th feature group data to the target prediction quantity, j is a hidden layer state of the i-th feature group data. i j

[0047] Optionally, the construction process of the BIGRU model includes:

[0048] ​​​​The initial BIGRU model is constructed by using a forward GRU network and a reverse GRU network.

[0049] The hyperparameters of the initial BIGRU model are optimized by using a multiverse algorithm to obtain a target BIGRU model, wherein the hyperparameters of the initial BIGRU model include a learning rate, an iteration number, a number of forward hidden layer nodes, and a number of backward hidden layer nodes.

[0050] Optionally, the process of optimizing the hyperparameters of the initial BIGRU model by using the multiverse algorithm includes:

[0051] In the population initialization process, a Circle chaotic mapping is added to generate a chaotic sequence to initialize the universes through the chaotic sequence.

[0052] The expansion rates of the universes are calculated, and an adaptive mechanism is introduced to adjust the adaptive adjustment mechanism of the wormhole existence probability E and the travel distance rate D parameter, and the universe expansion rate is updated.

[0053] A fitness function is constructed, the fitness is calculated, and it is determined whether the maximum iteration number is reached, if yes, the optimized hyperparameters are obtained.

[0054] Optionally, the formula of introducing the adaptive mechanism to adjust the adaptive adjustment mechanism of the wormhole existence probability E and the travel distance rate D parameter, and updating the universe expansion rate is:

[0055]

[0056] In the formula, represents the jth parameter of the current optimal universe, , , represents a random number subject to a uniform distribution and between [0, 1], represents the wormhole existence probability of each generation of universes, and D represents the travel distance rate, represents the minimum value of the jth parameter, represents the maximum value of the jth parameter, represents the jth parameter of the ith initialized universe.

[0057] Optionally, the fitness function formula is:

[0058]

[0059] In the formula, is the number of correctly classified samples; is the total number of samples.

[0060] In a second aspect, the embodiments of the present application also provide a state evaluation system of a transformer on-load tap changer, including:

[0061] An acquisition unit is configured to acquire electrical performance parameters and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatographic data, and the mechanical performance parameters include a vibration signal and a driving motor current signal;

[0062] A first generation unit is configured to generate an oil chromatographic data feature matrix for the oil chromatographic data, wherein the oil chromatographic data feature matrix includes a volume fraction of a gas generated by decomposition of insulation oil of the on-load tap changer in an abnormal situation, a total hydrocarbon value, a temperature of an overheating fault point of the on-load tap changer, a temperature of a discharge fault point, and an acetylene growth rate;

[0063] A second generation unit is configured to extract vibration signal features of the hydroelectric generator set vibration signal by using a multi-scale wave fluctuation spread entropy for the vibration signal, and introduce energy entropy, kurtosis, and spectral kurtosis to generate a vibration signal feature matrix;

[0064] A third generation unit is configured to generate a driving motor current signal feature matrix for the driving motor current signal, wherein the driving motor current signal feature matrix includes a duration of the driving motor current and a sum of absolute values of the driving motor current in a motor starting stage, a motor stable running stage, and a motor stopping running stage;

[0065] A fusion unit is configured to fuse the oil chromatographic data feature matrix, the vibration signal feature matrix, and the driving motor current signal feature matrix to generate a multi-dimensional information feature matrix;

[0066] An evaluation unit is configured to calculate correlations between data in the multi-dimensional information feature matrix by using an attention mechanism, assign weights to the multi-dimensional information feature matrix, obtain data samples with the weights, and input the data samples with the weights into a pre-constructed BIGRU model to realize state evaluation of the on-load tap changer of the transformer.

[0067] In a third aspect, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements steps of the state evaluation method of the on-load tap changer of the transformer when executing the program.

[0068] In a fourth aspect, an embodiment of the present application further provides a storage medium having a computer program stored thereon, and the computer program is executable on a processor to implement steps of the state evaluation method of the on-load tap changer of the transformer.

[0069] As can be seen from the above technical solutions, the present application has the following advantages:

[0070] The transformer on-load tap changer state evaluation method, system, electronic device and medium provided by the application comprehensively select feature data capable of reflecting electrical characteristics and mechanical characteristics of the on-load tap changer, form a feature matrix, and utilize the transformer on-load tap changer state evaluation model based on multi-dimensional information fusion to perform fault diagnosis, thereby improving the diagnostic accuracy and adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0071] In order to more clearly illustrate the technical solutions of the present application, the drawings required to be used in the description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor on the basis of these drawings.

[0072] Figure 1 A flowchart of a transformer on-load tap changer state evaluation method provided by an embodiment of the present application;

[0073] Figure 2 A timing diagram of an on-load tap changer vibration signal provided by an embodiment of the present application;

[0074] Figure 3 A coarse-grained time series diagram provided by an embodiment of the present application;

[0075] Figure 4 A flowchart of an MFDE method provided by an embodiment of the present application;

[0076] Figure 5 A motor current waveform division result schematic diagram provided by an embodiment of the present application;

[0077] Figure 6 A structure diagram of an attention mechanism provided by an embodiment of the present application;

[0078] Figure 7 A structure diagram of a GRU provided by an embodiment of the present application;

[0079] Figure 8 A structure diagram of a BIGRU provided by an embodiment of the present application;

[0080] Figure 9 An IMVO algorithm optimization flowchart provided by an embodiment of the present application;

[0081] Figure 10 A detailed flowchart of a transformer on-load tap changer state evaluation method provided by an embodiment of the present application;

[0082] Figure 11 A structure schematic diagram of a transformer on-load tap changer state evaluation system provided by an embodiment of the present application;

[0083] Figure 12 Fig. 1 shows a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0084] Various embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. The present disclosure can have various embodiments, and adjustments and changes can be made therein. However, it should be understood that there is no intention to limit various embodiments of the present disclosure to the specific embodiments disclosed herein, but the present disclosure should be understood to encompass all adjustments, equivalents, and / or alternatives falling within the spirit and scope of various embodiments of the present disclosure.

[0085] Hereinafter, the term "include" or "may include" used in various embodiments of the present disclosure indicates the presence of the disclosed function or operation, and does not limit the addition of one or more functions or operations. In addition, as used in various embodiments of the present disclosure, the terms "include", "have", and their conjugates merely mean to indicate a specific feature, number, step, operation, or combination of the foregoing, and should not be understood as first excluding the presence or possibility of addition of one or more other features, numbers, steps, operations, or combinations of the foregoing.

[0086] In various embodiments of the present disclosure, the expression "or" or "at least one of A or / and B" includes any combination of the listed terms or all combinations thereof. For example, the expression "A or B" or "at least one of A or / and B" can include A, can include B, or can include both A and B.

[0087] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only some of the embodiments of the present disclosure, not all the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by a person of ordinary skill in the art without creative labor fall within the scope of protection of the present disclosure.

[0088] Referring to Figure 1 Fig. 1 shows a schematic diagram of a hardware structure of an electronic device according to an embodiment of the present disclosure.

[0089] Step 100: Obtain electrical performance parameters and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatogram data, and the mechanical performance parameters include vibration signals and driving motor current signals.

[0090] Specifically, the performance indicators of OLTC mainly include electrical performance and mechanical performance. In the evaluation of electrical performance, when the transformer vacuum on-load tap changer is in normal operation, arc extinguishing in the switching process is generally carried out in the vacuum tube, the insulation oil of the tap changer does not directly undertake arc extinguishing, and the insulation oil will not deteriorate and produce gas. When the vacuum on-load tap changer has an electrical fault, characteristic gas will be produced in the insulation oil. Oil chromatographic analysis can effectively find the latent electrical fault existing in the vacuum on-load tap changer; in the evaluation of mechanical performance, on the one hand, the vibration signal generated by the collision of moving and static contacts in the OLTC switching process contains the mechanical state information of each component, and the vibration analysis method can effectively evaluate the mechanical state of the OLTC. On the other hand, the OLTC is driven by a driving motor to provide driving force for spring energy storage, and through a series of mechanical actions of components such as connecting rods, gearboxes, quick mechanisms and contacts, the gear switching is completed, and the motor current signal also contains rich mechanical state information. In summary, the oil chromatographic data, vibration signal and driving motor current signal of the on-load tap changer are selected.

[0091] Step 101: generating an oil chromatographic data feature matrix for the oil chromatographic data, wherein the oil chromatographic data feature matrix includes the volume fraction of gas produced by decomposition of the insulation oil of the on-load tap changer under abnormal conditions, the total hydrocarbon value, the temperature of the overheat fault point of the on-load tap changer, the temperature of the discharge fault point and the acetylene growth rate.

[0092] In some embodiments, the insulation oil of the on-load tap changer will not decompose to produce gas under normal conditions, and when the arc of the disconnecting switch in series with the vacuum tube cannot be extinguished due to vacuum tube leakage or insulation performance degradation, gas will be produced by decomposition. This embodiment selects five characteristic gases As the main research object, the defect conditions corresponding to the contents of different characteristic gases are analyzed and calculated. Therefore, the first introduced is The volume fractions of the five gases and the total hydrocarbon value are taken as characteristic values. It should be noted that for the same type of switch, due to different manufacturers, different models and different structural designs, the production of characteristic gases under defect conditions will also be very different.

[0093]

[0094] In the formula, is the volume fraction of gas i, 10 -6 ; is the volume of gas i, the unit is consistent with the total volume; is the total volume of the mixed gas.

[0095] Specifically, the gas produced by decomposition of the insulation oil of the on-load tap changer under abnormal conditions at least includes hydrogen, methane, ethane, ethylene and acetylene.

[0096] In some embodiments, the overheat fault of the on-load tap changer is generally caused by poor contact of current-carrying contacts, long-term overload or poor heat dissipation of the medium around the contacts, and the like. The overheat of the contacts inevitably causes the decomposition of insulating oil, and H2 is usually the first characteristic gas to appear. As the temperature rises, CH4, C2H6 and C2H4 will be produced, at which time the fault characteristic gases are CH4 and C2H4, and as the temperature of the fault point rises, the proportion of C2H4 gradually increases. The temperature of the overheat fault point of the on-load tap changer is calculated according to the following formula:

[0097]

[0098] In the formula, T G is the overheat fault point temperature, is the volume fraction of ethylene, is the volume fraction of ethane.

[0099] The electrical fault of the transformer vacuum on-load tap changer is mainly discharge fault. The on-load tap changer discharge fault can be divided into three types according to the energy density of discharge, i.e. partial discharge, spark discharge and arc discharge. Partial discharge and spark discharge belong to low-energy discharge fault, and arc discharge is a malignant fault of high-energy discharge. It is generally believed that partial discharge and spark discharge will not cause insulation breakdown of the vacuum on-load tap changer soon, which is mainly reflected in abnormal oil chromatographic analysis, light gas action of the on-load tap changer, and the fault is relatively easy to find and handle. Arc discharge usually leads to heavy gas action of the on-load tap changer and transformer trip. H2 and C2H2 are usually taken as the main characteristic gases of discharge fault, and arc discharge means that the temperature of the fault part is extremely high. The temperature of the discharge fault point is calculated according to the following formula:

[0100]

[0101] In the formula, is the discharge fault point temperature, is the volume fraction of hydrogen, is the volume fraction of acetylene.

[0102] At the same time, in actual operation, acetylene and other discharge characteristic gases and ethylene, methane and other overheat characteristic gases are often detected in the insulating oil of the normally operating vacuum on-load tap changer due to the main contact recovery voltage, transition resistance wire heating and the like, but the acetylene gas production rate is generally slow. Therefore, after the characteristic gases are detected, in order to effectively distinguish whether the gas production of the on-load tap changer is normal or fault, the acetylene growth rate is introduced as a new characteristic value in the present application. The state of the on-load tap changer is comprehensively judged in combination with the volume fraction content of different characteristic gases of the on-load tap changer, the fault point temperature and the acetylene growth rate. The acetylene growth rate is calculated according to the following formula:

[0103]

[0104] wherein, is the acetylene growth rate, is the change in acetylene volume fraction, is the time interval.

[0105] In summary, a feature matrix is formed for the oil chromatography data:

[0106] .

[0107] Step 102: For the vibration signal, a multi-scale fluctuation dispersion entropy is used to extract the vibration signal features of the hydroelectric unit, and energy entropy, kurtosis and spectral kurtosis are introduced to generate a vibration signal feature matrix.

[0108] In recent years, a mechanical state vibration detection method has been proposed, which collects vibration signals by arranging vibration sensors on the surface of the equipment, and realizes state detection by extracting the features of the vibration signals under different working conditions. Since the vibration signals and their related features can directly reflect the internal faults of the equipment, the vibration detection method has high sensitivity and accuracy for mechanical state detection. At the same time, since the entire measurement is non-intrusive, the detection is easy to operate and has strong engineering practical value. However, the fault vibration signal contains a large amount of noise interference, and its dynamics generally has strong nonlinearity, so traditional nonlinear system analysis indicators such as mean, variance, kurtosis, skewness, peak value, waveform factor, etc. It is difficult to effectively extract fault feature information, so new indicators need to be adopted to analyze the systematic properties or dynamics of the signal. The present application introduces multi-scale fluctuation dispersion entropy (MFDE) to characterize the complexity of time series at different scales for the vibration signal of the on-load tap changer, so as to strengthen the fault features of the vibration signal. The time series diagram of the on-load tap changer vibration signal is shown in the accompanying Figure 2 .

[0109] Specifically, the calculation process of the multi-scale fluctuation dispersion entropy includes the following steps:

[0110] S1: Map the coarse-grained time series, and linearly transform the mapping result to obtain Z.

[0111] For example, the coarse-grained time series is mapped to , that is,

[0112] ;

[0113] Wherein, μ and are the expectation and variance of x, respectively.

[0114] Then linearly transform to , The middle element is an integer:

[0115]

[0116] where round() is the rounding method, and c is the number of classes in the symbolization process.

[0117] S2: Calculate the embedding vectors according to the embedding dimension and the time delay, and map each embedding vector to a fluctuation-based diffusion pattern.

[0118] For example, the embedding vector is calculated according to the embedding dimension m and the time delay d :

[0119]

[0120] Map each embedding vector to a fluctuation-based diffusion pattern :

[0121]

[0122] The number of all possible diffusion patterns of each is .

[0123] S3: Calculate the probability of each diffusion pattern, and calculate the fluctuation-based diffusion entropy based on the Shannon entropy calculation method.

[0124] For example, the probability of each possible diffusion pattern is calculated as :

[0125]

[0126] where count() represents the number of mapping to, i.e., the probability is equal to the number of mapping to divided by the total number of embedding vectors corresponding to the embedding dimension m.

[0127] S4: Calculate the fluctuation-based diffusion entropy under the scale factor, and obtain the multi-scale fluctuation diffusion entropy function.

[0128] For example, based on the Shannon entropy calculation method, the fluctuation-based diffusion entropy is calculated as

[0129]

[0130] ​The difference between adjacent elements of the scatter pattern is considered in FDE calculation, which is called wave-based scatter pattern. In this algorithm, a pattern vector with dimension m-1 can be obtained, each element of the pattern vector ranges from -c+1 to c-1, so there are wave-based scatter patterns. When all scatter patterns have equal probability values, the entropy value is maximum, which is , at this time the signal is completely random.

[0131] In MFDE calculation, refer to Figure 3 , Figure 4 , for a given time series , first divide it into non-overlapping segments with length by using the floor operation, τ is the scale factor, is the floor operation. Then the average value of each segment is calculated. The coarse-grained process of MFDE can be simply understood as averaging the original time series in a window with length τ to obtain the time series:

[0132]

[0133] Calculate FDE under the scale factor τ to obtain MFDE as a function of scale factor τ:

[0134]

[0135] The parameters of MFDE are embedding dimension m, class number n, time delay s, and scale factor τ. If the embedding dimension m is too large, it will be difficult to detect small changes in the signal, and if m is too small, it will be difficult to observe the dynamic changes in the signal. The usual selection range is 2≤m≤5, and m is usually taken as 3; the class number is usually taken as 3≤n≤8, and n is taken as 3; the time delay s is generally set to 1 to avoid losing important frequency information due to too large time delay. When τ=1, the coarse-grained time series is equal to the original signal, Figure 3 shows the coarse-grained time series when τ=2 and τ=3, Figure 4 illustrates the flowchart of the MFDE method, where τ m is the maximum scale factor set.

[0136] Meanwhile, the application still calculates some traditional time-frequency domain feature values to enhance the fault representation capability, mainly extracting three types of feature values: energy entropy, kurtosis, and spectral kurtosis.

[0137] In some embodiments, energy entropy describes the uncertainty of energy distribution in different frequency bands or time windows, and can be used to measure the complexity and randomness of the signal. The energy entropy is calculated according to the following formula:

[0138]

[0139] In the formula, represents the amplitude of vibration, represents the ratio of the current position energy to the total energy, represents the time series.

[0140] In some embodiments, the kurtosis is a measure of the fourth-order standard deviation, which is used to describe the degree of signal spikes, and can be used to capture non-periodic mutations in the signal, and is used to detect transient faults in fault diagnosis. The formula is as follows:

[0141]

[0142] In the formula, N is the data sample capacity, is the i-th data sample, is the sample mean.

[0143] Kurtosis is a feature sensitive to outliers, and a single extreme value will affect the kurtosis of the entire section. Spectral kurtosis is an extension of the time-domain feature kurtosis, and can be used to extract transient vibration frequency bands and envelope feature analysis. Spectral kurtosis can represent the sharpness of the frequency spectrum, and can be approximated as the kurtosis of the power spectral density. The envelope spectrum of the signal is obtained by using the fast Fourier transform on the envelope of the signal, and the square of the frequency spectrum is normalized to approximate the power spectral density of the signal. The power spectral density is calculated as follows:

[0144]

[0145] In the formula, represents the frequency, represents the power spectral density of the current frequency, and the spectral kurtosis is calculated according to the following formula:

[0146] ;

[0147] In the formula, represents the frequency, represents the power spectral density of the current frequency, represents the average power spectral density.

[0148] Step 103: generating a driving motor current signal feature matrix for the driving motor current signal, wherein the driving motor current signal feature matrix includes the duration of the driving motor current and the sum of the absolute values of the driving motor current in the motor starting stage, the motor stable running stage and the motor stopping stage.

[0149] In some embodiments, the driving motor is a power source of the on-load tap changer, which adjusts the working position of the on-load tap changer to the required position during on-load voltage regulation. By observing the collected current waveform, it can be found that the motor current waveform can be divided into three stages:

[0150] Phase 1: motor starting phase. Its characteristic is that there is a large amplitude inrush current at the moment of motor starting with load, and the current signal is stable after about 0.3s;

[0151] Phase 2: motor stable operation phase. Its characteristic is that the motor works stably, the amplitude of the current signal is basically unchanged, and the OLTC switching action is completed in this phase;

[0152] Phase 3: motor stopping phase. A period of time after the OLTC switching action is completed, the motor current is cut off and drops to 0.

[0153] The driving motor is the power source of the OLTC operation, and its output torque is closely related to the current signal. The duration T c of the driving motor current is equivalent to the total switching time of the on-load tap changer, so the application takes Tc as the first feature of the driving motor current signal.

[0154] When the on-load tap changer has faults such as jamming and spring fatigue, the driving motor torque will change. The common on-load tap changer driving motor is a three-phase asynchronous motor. In the following, the relationship between the output torque T2 of the driving motor and the stator current I1 will be discussed taking the three-phase asynchronous motor as an example. According to the power balance relationship, when the three-phase asynchronous motor is in stable operation, the electromagnetic torque P em is related to the stator phase current I1 as follows:

[0155]

[0156] In the formula: P1 is the input power; P Cu1 is the stator copper loss; P Fe is the stator iron loss, which is ignored here; m1 is the number of phases; U1 is the stator phase voltage; is the stator power factor; R1 is the equivalent resistance on the stator side.

[0157] Ignoring mechanical loss and additional loss, the output power P2 of the three-phase asynchronous motor is:

[0158]

[0159]

[0160] In the formula: P m is the mechanical power of the motor; s is the slip of the motor; n1 is the synchronous speed; n is the rotor speed.

[0161] From the power torque relationship, the output torque T2 of the three-phase asynchronous motor is:

[0162] ; ;

[0163] Wherein, Ω is the mechanical angular velocity of the rotor.

[0164] Combining the above formula with the formula can be obtained:

[0165]

[0166] Since , and n1, U1, is a constant value at steady state, the output torque T2 of the three-phase asynchronous motor is approximately proportional to the stator current I1. Therefore, the sum of the absolute values of the driving motor current is taken as a new feature of the driving motor current signal in the present application. At the same time, when different mechanical faults occur in the OLTC, the current feature parameters of each stage change differently, so it is necessary to divide the three stages. Based on the foregoing analysis, the intervals of the three stages are: , , The division result is shown in the accompanying Figure 5 Therefore, the sum of the absolute values of each segment is calculated as a feature value.

[0167]

[0168] In the formula, is a stage of the driving motor current signal. The three stages are represented by I, II, and III.

[0169] Step 104: Fusing the oil chromatographic data feature matrix, the vibration signal feature matrix, and the driving motor current signal feature matrix to generate a multi-dimensional information feature matrix.

[0170] As described above, a feature matrix is formed for the driving motor current signal: .

[0171] Through the feature analysis of the above-mentioned on-load tap changer oil chromatographic data, vibration signal, and driving motor current signal, a large multi-dimensional information feature matrix is finally formed.

[0172] In some embodiments, the orders of magnitude of different feature data differ greatly. If such data with a large difference in orders of magnitude is directly used as input of the evaluation model, it is easy to cause small data to be swallowed by the network in a case of approximately equal to zero relative to large data, and the input data with a lower order of magnitude is equivalent to zero input by default. However, most of the time, these data with small values are crucial for network training. Therefore, the different feature value data of the oil chromatographic data, vibration signal, and driving motor current signal obtained in the foregoing are subjected to Min-Max normalization processing to obtain data between 0 and 1, reducing the interference of the orders of magnitude of the data.

[0173] Step 105: Calculate the correlation between data in the multidimensional information feature matrix using the attention mechanism and assign weights to the multidimensional information feature matrix to obtain weighted data samples. Then, input the weighted data samples into the pre-built BIGRU model to realize the state assessment of the on-load tap changer of the transformer.

[0174] To better identify useful information between input feature data and the target output, highlight key features relevant to fault diagnosis, and improve fault diagnosis accuracy, this application applies the Attention Mechanism (AM) to multidimensional feature data analysis. It calculates the correlation between data and assigns different weights to different feature data, thus replacing the original data. The principle of the Attention Mechanism is explained below. Essentially, the Attention Mechanism is a weighted probability distribution mechanism that can quickly filter out high-value information from a large amount of data, assigning high weights to important content and correspondingly low weights to less important content. The structure of the Attention Mechanism can be viewed as two modules: Encode and Decode. Encode is the encoder, transforming input data into binary data; Decode is the decoder, converting binary data into the output data type. Its structure is shown in the attached figure. Figure 6 In the picture X 1— X 10 Indicates the input data type.

[0175] Specifically, the correlation between data in the multidimensional information feature matrix is ​​calculated using an attention mechanism, and weights are assigned to the multidimensional information feature matrix to obtain weighted data samples, including an additive attention scoring mechanism:

[0176]

[0177] In the formula, This is the attention scoring mechanism after the hidden layer has undergone one fully connected operation. u s , w , b These are the randomly initialized time series, attention weight matrix, and bias term matrix, respectively. for i Type of feature j The hidden layer state of the group data.

[0178] Specifically, the attention mechanism is used to calculate the correlation between data in the multidimensional information feature matrix and assign weights to the multidimensional information feature matrix to obtain weighted data samples. It also includes an attention distribution mechanism.

[0179]

[0180] In the formula, for i Type of featurej The degree of attention of the group data to the target prediction quantity.

[0181] The correlation between data in the multi-dimensional information feature matrix is calculated by using the attention mechanism, and a weight is assigned to the multi-dimensional information feature matrix to obtain a data sample with a weight, and the information weighted summation mechanism is further included:

[0182]

[0183] In the formula, is a score vector after weighted summation, i A feature of the first j The degree of attention of the group data to the target prediction quantity, i A feature of the first j The hidden layer state of the group data.

[0184] In a specific embodiment, the construction process of the BIGRU model includes:

[0185] An initial BIGRU model is constructed by using a forward GRU network and a reverse GRU network;

[0186] The hyperparameters of the initial BIGRU model are optimized by using a multiverse algorithm to obtain a target BIGRU model, wherein the hyperparameters of the initial BIGRU model include a learning rate, an iteration number, a forward hidden layer node number, and a reverse hidden layer node number.

[0187] In some embodiments, the structural diagram of the GRU is as shown in the accompanying Figure 7 The forgetting gate and the input gate of the LSTM network are replaced by an update gate in the GRU, and the forgetting gate and the input gate are combined. Figure 7 In the formula, x t is the input at the current time (i.e. t time); h t-1 and h t and respectively are the state outputs of the GRU at the previous time (i.e. is the candidate state at the current time; r t and z t respectively are the outputs of the reset gate and the update gate at the current time. The output calculation formula of the GRU is as follows.

[0188]

[0189] ​​

[0190] wherein: W r and b r are the weights and biases of the reset gate, respectively; W z and b z are the weights and biases of the update gate, respectively; W d and b d are the weights and biases of the gate unit, respectively; and

[0191] BIGRU is an improvement of GRU, which contains two opposite directions of GRU (i.e. forward GRU and reverse GRU), and the two GRUs do not interfere with each other, both as input and together control the output, as shown in the accompanying drawings. BIGRU can mine more information from the monitoring data than GRU, so the algorithm has the characteristics of fewer parameters and strong optimization ability. BiGRU network is jointly constructed by forward GRU network and reverse GRU network, and the two layers of GRU network are responsible for capturing historical information and future information, respectively. Finally, the outputs of the two networks are integrated according to their respective positions, which can improve the memory ability and prediction accuracy. The super parameter optimization of the BIGRU model mainly includes: learning rate (LR), iteration number (NI), forward hidden layer node number (NF), and backward hidden layer node number (NB) a total of four parameters. Figure 8

[0192] The present application introduces a multi-universe algorithm (MVO) to optimize the super parameters of the BIGRU model. MVO algorithm is a meta-heuristic algorithm based on the physical motion law among universes, which has good global optimization ability. In the cyclic model of the multi-universe theory, some physicists believe that white holes are produced at the place where parallel universes collide, and white holes repel all objects including light beams with extremely high repulsive force. The black hole, which is often observed, behaves completely opposite to the white whole, and the black hole attracts all objects including light beams with extremely high attractive force. Wormholes connect different universes together and play the role of time-space travel tunnels in algorithm theory, so that objects can instantly travel from one universe to another. Next, the MVO algorithm will be described in detail.

[0193] MVO algorithm assumes that each set of candidate solutions of the problem is a universe, and each parameter in the candidate solution is an object in the universe. The jth parameter of the ith universe is wherein represents the minimum value of the jth parameter, ​represents the maximum value of the jth parameter. It is assumed that there are I universes, each of which contains J parameters, and each parameter in the universe is initialized according to a random number:

[0194]

[0195] wherein, represents the ith universe, represents the jth parameter of the ith initialized universe.

[0196] The MVO algorithm considers that each universe has an expansion rate representing the expansion degree of the universe, the expansion rate is proportional to the corresponding fitness function value of the solution, and the expansion rate determines the formation of black holes, white holes and wormholes.

[0197] The expansion rate of each universe can be obtained according to the objective function f , as shown in the following formula:

[0198]

[0199] Then, the expansion rate of each universe is standardized to obtain the standardized universe expansion rate , as shown in the following formula:

[0200]

[0201] The MVO algorithm is divided into an exploration stage and a development stage. In the exploration stage, the standardized expansion rate of the universe is first calculated to calculate the probability of generating a white hole for each universe , as shown below:

[0202]

[0203] Then, the cumulative probability of generating a white hole for each universe is calculated , as shown below:

[0204]

[0205] The exploration stage updates the parameters of the black hole and white hole based on the roulette mechanism, as shown in the following formula:

[0206]

[0207] wherein, represents the jth parameter of the kth universe generating a white hole, represents a random number subject to a uniform distribution and between [0, 1].

[0208] The development stage uses a wormhole for fine local search, and the development stage updates the parameters of the wormhole based on the wormhole, as shown in the following formula:

[0209]

[0210] in, Let D represent the j-th parameter of the current optimal universe, D represent the travel distance rate, and E represent the probability of the wormhole existing. , , It represents a random number that follows a uniform distribution and lies between [0, 1].

[0211] Two important coefficients exist during the development phase: the probability of a wormhole's existence (E) and the travel distance (D). The probability of a wormhole's existence is shown in the formula:

[0212]

[0213] in, This represents the minimum probability of a wormhole existing (0.2 in the MVO algorithm). q represents the maximum probability of a wormhole's existence (1 for the MVO algorithm), q represents the current iteration number, and Q represents the maximum iteration number.

[0214] The travel distance rate is shown in the formula:

[0215]

[0216] h is an adjustment parameter for D. The higher the h, the faster the local search. In this embodiment, h is set to 8.

[0217] In the MVO algorithm, the optimization process begins by initializing I random universes. Then, in each iteration, objects in universes with high expansion rates tend to move towards universes with low expansion rates via black hole-white hole tunnels. Simultaneously, each universe may randomly teleport to the current optimal universe via wormholes. This process iterates until the final iteration stopping condition is met, typically a predefined maximum number of iterations.

[0218] For example, the traditional MVO algorithm uses a random generation method to generate the initial population, resulting in poor particle distribution uniformity. Therefore, see [link to relevant documentation]. Figure 9 As shown, this application incorporates Circle chaotic mapping to generate chaotic sequences during population initialization. Using these chaotic sequences to initialize the universe improves the optimization capability and search efficiency of the MVO algorithm. The process of optimizing the hyperparameters of the initial BIGRU model using the multiverse algorithm includes the following steps:

[0219] S1: During the initialization process, a Circle chaotic mapping is added to generate a chaotic sequence in order to initialize the universe through the chaotic sequence.

[0220] For example, the mathematical representation of the Circle chaotic map is shown below:

[0221]

[0222] where mod denotes the modulo operation, represents the generated chaotic sequence of the jth parameter of the ith universe.

[0223] Compared with random numbers, the Circle mapping can generate chaotic sequences with uniform distribution, avoiding the possibility of generating blank spaces. Using the Circle mapping for universe initialization can provide a high-quality search space for the algorithm, which is beneficial to improve the convergence accuracy of the algorithm.

[0224] S2: Calculate the expansion rate of each universe, and introduce an adaptive mechanism to adjust the wormhole existence probability E and the adaptive adjustment mechanism of the travel distance rate D parameter, and update the universe expansion rate.

[0225] Exemplary, based on the above analysis of the MVO algorithm, it can be known that in the algorithm, adjusting the two important parameters of wormhole existence probability E and travel distance rate D can play a key role in balancing the development and search stages. The imbalance of the two stages in the traditional multi-universe optimization algorithm can lead the algorithm to fall into a local optimal solution or even fail. In order to ensure the effectiveness of the algorithm, the present application adjusts E and D using an adaptive mechanism based on the conventional algorithm, so that the values of the two parameters are automatically adjusted with the changes of fitness and iteration number, making the optimization effect of the algorithm better. The mathematical expression of the adaptive adjustment mechanism of E is as follows:

[0226]

[0227] In the formula, represents the wormhole existence probability of each generation of universe; , and respectively represent the average value, maximum value and minimum value of the fitness function value of each generation of universe in the iteration process; represents the fitness function value of the ith generation of universe object.

[0228] The adaptive adjustment mechanism of D can be represented by the following formula:

[0229]

[0230] Specifically, the adaptive adjustment mechanism of the wormhole existence probability E and the adaptive adjustment mechanism of the travel distance rate D parameter are introduced, and the formula for updating the universe expansion rate is:

[0231]

[0232] In the formula, represents the jth parameter of the current optimal universe, , , represents a random number subject to uniform distribution and between [0, 1], represents the probability of wormhole existence of each generation universe, and D represents the distance rate, represents the minimum value of the jth parameter, represents the maximum value of the jth parameter, represents the jth parameter of the ith initialized universe.

[0233] S3: Construct the fitness function, calculate the fitness, and judge whether the maximum iteration number is reached. If yes, the optimized hyperparameters are obtained.

[0234] Specifically, the IMVO algorithm is used to optimize four parameters of the BIGRU model, i.e., learning rate (LR), iteration number (NI), forward hidden layer node number (NF), and backward hidden layer node number (NB). The fitness function formula is:

[0235]

[0236] In the formula, is the number of correctly classified samples; is the total number of samples.

[0237] In an embodiment, Figure 10 is a detailed flowchart of a transformer on-load tap changer state evaluation method provided by an embodiment of the application. The embodiment is further optimized and expanded on the basis of the above embodiments.

[0238] Step 1: Oil chromatographic data feature extraction. Select Five characteristic gases are selected as the main research objects. The volume fraction content, fault point temperature, and acetylene growth rate of different characteristic gases of the on-load tap changer form an oil chromatographic data feature matrix (Xoil), and the state of the tap changer is comprehensively judged.

[0239] Step 2: On-load tap changer vibration signal feature extraction. The multiscale fluctuation dispersion entropy (MFDE) is introduced to represent the complexity degree of the time series at different scales, so as to strengthen the fault features of the vibration signal. At the same time, three types of traditional time-frequency domain characteristic values, energy entropy, kurtosis, and spectral kurtosis, are calculated to enhance the fault representation ability, and a vibration signal feature matrix is formed.

[0240] Step 3: Feature extraction of the driving motor current signal of the on-load tap changer. First, the duration Tc of the driving motor current is taken as the first feature of the driving motor current signal. Second, according to the on-load tap changer change process, the driving motor current signal is divided into three stages, and the sum of the absolute values of the driving motor current in different stages is taken as a new feature of the driving motor current signal. Thus, the driving motor current signal feature matrix is formed.

[0241] Step 4: Data normalization processing. The obtained data samples are subjected to Min-Max normalization processing.

[0242] Step 5: Attention mechanism correlation calculation. The correlation between data is calculated by using the attention mechanism (AM), and different weights are given to different feature data, which replaces the original data to obtain data samples with specific weights.

[0243] Step 6: BIGRU network model related parameter optimization. The key parameters affecting the accuracy of BIGRU fault diagnosis, such as learning rate, iteration number, forward hidden layer node number, and backward hidden layer node number, are optimized by using IMVO to obtain the optimal parameters of the BIGRU diagnosis model.

[0244] Step 7: Model training and on-load tap changer state evaluation. The data samples with specific weights are input into the BIGRU model, and the BIGRU network model under the optimal parameter group is used to train and analyze the data, so as to realize the state evaluation of the on-load tap changer based on multi-dimensional information parameters and realize multi-dimensional information fusion diagnosis.

[0245] It should be understood that the sequence numbers of the steps in the above embodiments do not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0246] As shown in Figure 11 The following is an embodiment of the state evaluation system of the on-load tap changer of the transformer provided by the embodiment of the present disclosure, which belongs to the same inventive concept as the state evaluation method of the on-load tap changer of the transformer described above. The details not described in detail in the embodiment of the new energy power supply photovoltaic panel power generation and electricity storage data management and control system can be referred to the embodiment of the state evaluation method of the on-load tap changer of the transformer.

[0247] The acquisition unit 110 is configured to acquire electrical performance parameters and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatographic data, and the mechanical performance parameters include vibration signals and driving motor current signals.

[0248] The first generation unit 111 is configured to generate an oil chromatographic data feature matrix for the oil chromatographic data, wherein the oil chromatographic data feature matrix comprises a volume fraction of a gas generated by an insulation oil of the on-load tap changer in an abnormal situation, a total hydrocarbon value, a temperature of an overheating fault point of the on-load tap changer, a temperature of a discharge fault point, and an acetylene growth rate.

[0249] The second generation unit 112 is configured to extract vibration signal features of the hydroelectric generator set by using a multi-scale wave fluctuation spread entropy for the vibration signal, and introduce energy entropy, kurtosis, and spectral kurtosis to generate a vibration signal feature matrix.

[0250] The third generation unit 113 is configured to generate a driving motor current signal feature matrix for the driving motor current signal, wherein the driving motor current signal feature matrix comprises a duration of the driving motor current and a sum of absolute values of the driving motor current in a motor starting stage, a motor stable running stage, and a motor stopping running stage.

[0251] The fusion unit 114 is configured to fuse the oil chromatographic data feature matrix, the vibration signal feature matrix, and the driving motor current signal feature matrix to generate a multi-dimensional information feature matrix.

[0252] The evaluation unit 115 is configured to calculate correlations between data in the multi-dimensional information feature matrix by using an attention mechanism, assign weights to the multi-dimensional information feature matrix, obtain data samples with the weights, and input the data samples with the weights into a pre-constructed BIGRU model to realize state evaluation of the on-load tap changer of the transformer.

[0253] Figure 12 FIG. 1 is a hardware structure schematic diagram of an electronic device for implementing various embodiments of the present application.

[0254] The state evaluation method of the on-load tap changer of the transformer provided by the embodiments of the present application can be applied to an electronic device. Those skilled in the art can understand that the electronic device structure involved in the embodiments of the present application does not constitute a limitation on the electronic device, and the electronic device can include more or fewer components than the diagram, or combine certain components, or different component arrangements. In the embodiments of the present application, the electronic device includes but is not limited to a laptop computer, a desktop computer, a workstation, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the embodiments of the present application described and / or claimed herein.

[0255] The electronic device can include a processor, an external memory interface, an internal memory, a universal serial bus (USB) interface, a charging management module, a power management module, a battery, a wireless communication module, an audio module, a speaker, a microphone, a sensor module, a key, a camera, a display screen, and a SIM card interface, and the like.

[0256] It can be understood that the structure illustrated in the embodiments of the present application does not constitute a specific limitation on the electronic device. In other embodiments of the present application, the electronic device can include more or fewer components than the illustration, or combine certain components, or split certain components, or different arrangement of components. The illustrated components can be implemented in hardware, software, or a combination of software and hardware.

[0257] The processor can include one or more processing units, such as: the processor can include a central processing unit (CPU), an application processor (AP), a modem processor, a graphics processing unit (GPU), an image signal processor (ISP), a controller, a memory, a video codec, a digital signal processor (DSP), a baseband processor, and / or a neural-network processing unit (NPU), and the like. Among them, different processing units can be independent devices, or can be integrated in one or more processors.

[0258] Among them, the processor can be the nerve center and command center of the electronic device. The controller can generate operation control signals according to instruction operation codes and timing signals to complete the control of fetching instructions and executing instructions.

[0259] The memory in the processor can also be provided to store instructions and data. In some embodiments, the memory in the processor is a cache memory. The memory can save instructions or data that the processor has just used or repeatedly uses. If the processor needs to use the instructions or data again, it can directly call from the memory. Avoiding repeated access, reducing the waiting time of the processor, thus improving the efficiency of the system.

[0260] The external memory interface can be used to connect an external memory card, such as a MicroSD card, to expand the storage capacity of the electronic device. The external memory card communicates with the processor through the external memory interface to realize the data storage function. For example, save music, video, and other files in the external memory card.

[0261] The internal memory can be used to store computer executable program code including instructions. The processor performs various function applications and data processing of the electronic device by running the instructions stored in the internal memory. The internal memory can include a program area and a data area. The internal memory can include a high-speed random access memory, and can further include a non-volatile memory such as at least one disk storage device, a flash memory device, a universal flash storage (UFS), etc.

[0262] The wireless communication function of the electronic device can be implemented through an antenna, a wireless communication module, a modem processor, and a baseband processor, etc.

[0263] The wireless communication module can provide a wireless communication solution including wireless local area networks (WLAN) (such as a wireless fidelity (Wi-Fi) network), Bluetooth (BT), a global navigation satellite system (GNSS), frequency modulation (FM), near field communication (NFC), infrared technology (IR), etc. applied to the electronic device.

[0264] The electronic device can implement an audio function, etc. through an audio module, a speaker, a receiver, a microphone, an earphone interface, and an application processor, etc.

[0265] The electronic device can implement a photographing function through an ISP, a camera, a video codec, a GPU, a display screen, and an application processor, etc.

[0266] The electronic device can implement a display function through a GPU, a display screen, and an application processor, etc.

[0267] The GPU is a microprocessor for image processing, connected to the display screen and the application processor. The GPU is used to perform mathematical and geometric calculations for graphics rendering. The processor can include one or more GPUs that execute program instructions to generate or change display information.

[0268] The display screen is used to display images, videos, etc. The display screen includes a display panel.

[0269] In the storage medium provided in the present application, a program product capable of realizing a state evaluation method of a transformer on-load tap changer is stored.

[0270] In some possible implementation manners, the state evaluation method and system of the on-load tap changer of the transformer of the subject of the present disclosure can be implemented in the form of a program product, which includes program codes for causing a terminal device to perform the steps described in the above "Exemplary Method" section of the present specification according to various exemplary embodiments of the present disclosure when the program product is run on the terminal device.

[0271] The storage medium of the present disclosure can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium may, for example, be but is not limited to an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. More specific examples (non-exhaustive list) of the readable storage medium include an electrical connection having one or more wires, a portable disc, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0272] The above description of disclosed embodiments enables one of ordinary skill in the art to make or use the application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for assessing the condition of an on-load tap changer in a transformer, characterized in that, include: Obtain the electrical and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatography data and the mechanical performance parameters include vibration signals and drive motor current signals; For the oil chromatographic data, an oil chromatographic data feature matrix is ​​generated, wherein the oil chromatographic data feature matrix includes the volume fraction of gases produced by the decomposition of insulating oil of on-load tap changer under abnormal conditions, total hydrocarbon value, temperature of overheating fault point of on-load tap changer, temperature of discharge fault point, and acetylene growth rate. For the vibration signal, multi-scale wave walk entropy is used to extract the vibration signal features of the hydropower unit, and energy entropy, kurtosis and spectral kurtosis are introduced to generate a vibration signal feature matrix. For the drive motor current signal, a drive motor current signal feature matrix is ​​generated, wherein the drive motor current signal feature matrix includes the duration of the drive motor current and the sum of the absolute values ​​of the drive motor current during the motor start-up phase, the motor stable operation phase, and the motor stop operation phase. By integrating the oil chromatography data feature matrix, the vibration signal feature matrix, and the drive motor current signal feature matrix, a multidimensional information feature matrix is ​​generated. The correlation between data in the multidimensional information feature matrix is ​​calculated by using the attention mechanism and weights are assigned to the multidimensional information feature matrix to obtain weighted data samples. The weighted data samples are then input into a pre-built BIGRU model to achieve state assessment of the on-load tap changer of the transformer.

2. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, Under abnormal conditions, the gases produced by the decomposition of insulating oil in on-load tap changers include at least hydrogen, methane, ethane, ethylene, and acetylene.

3. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, Calculate the temperature at the overheating fault point of the on-load dismantling switch using the following formula: In the formula, T G The temperature at the point of overheating failure. This represents the volume fraction of ethylene. This represents the volume fraction of ethane.

4. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, The temperature at the discharge fault point can be calculated using the following formula: In the formula, The temperature at the discharge fault point. The integral number of hydrogen gas. This represents the volume fraction of acetylene.

5. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, The acetylene propagation rate can be calculated using the following formula: In the formula, For the acetylene growth rate, This represents the change in the volume fraction of acetylene. For time intervals.

6. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, The calculation process of multi-scale wave walk entropy includes: The coarse-grained time series is mapped, and the mapping result is linearly transformed to obtain Z; Embedding vectors are calculated based on embedding dimension and latency, and each embedding vector is mapped to a wave-based stalking pattern; Calculate the probability of each scattering model and, based on the Shannon entropy calculation method, calculate the wave-based scattering entropy; The wave-based spread entropy under the scale factor is calculated to obtain the multi-scale wave spread entropy function.

7. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, Calculate the energy entropy using the following formula: In the formula, Indicates the vibration amplitude. This represents the ratio of the current energy to the total energy. Represents a time series.

8. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, Calculate kurtosis using the following formula: In the formula, N is the data sample size. For the i-th data sample, This is the sample mean.

9. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, Calculate the spectral kurtosis using the following formula: ； ； In the formula, Indicates frequency, This represents the power spectral density at the current frequency. This represents the average power spectral density.

10. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, The correlation between data in a multidimensional information feature matrix is ​​calculated using an attention mechanism, and weights are assigned to the multidimensional information feature matrix to obtain weighted data samples. This includes an additive attention scoring mechanism. In the formula, This is the attention scoring mechanism after the hidden layer has undergone one fully connected operation. u s , w , b These are the randomly initialized time series, attention weight matrix, and bias term matrix, respectively. for i Type of feature j The hidden layer state of the group data.

11. The method for assessing the condition of an on-load tap changer of a transformer according to claim 10, characterized in that, The attention mechanism is used to calculate the correlation between data in the multidimensional information feature matrix and assign weights to the multidimensional information feature matrix to obtain weighted data samples. It also includes an attention distribution mechanism. In the formula, for i Type of feature j The degree of attention given to the target prediction quantity by the set of data.

12. The method for assessing the condition of an on-load tap changer of a transformer according to claim 11, characterized in that, The algorithm utilizes an attention mechanism to calculate the correlation between data in a multidimensional information feature matrix and assigns weights to the matrix, obtaining weighted data samples. It also includes an information-weighted summation mechanism. In the formula, This is the weighted sum of the fractional vectors. for i Type of feature j The degree of attention given to the target predicted quantity by the set of data. for i Type of feature j The hidden layer state of the group data.

13. The method for assessing the condition of an on-load tap changer of a transformer according to claim 1, characterized in that, The construction process of the BIGRU model includes: An initial BIGRU model is constructed using a forward GRU network and a reverse GRU network; The hyperparameters of the initial BIGRU model are optimized using the multiverse algorithm to obtain the target BIGRU model. The hyperparameters of the initial BIGRU model include the learning rate, the number of iterations, the number of forward hidden layer nodes, and the number of backward hidden layer nodes.

14. The method for assessing the condition of an on-load tap changer of a transformer according to claim 13, characterized in that, The process of optimizing the hyperparameters of the initial BIGRU model using the multiverse algorithm includes: During population initialization, a Circle chaotic map is added to generate a chaotic sequence in order to initialize the universe through the chaotic sequence; The expansion rate of each universe is calculated, and an adaptive adjustment mechanism is introduced to adjust the parameters of wormhole existence probability E and travel distance rate D, thereby updating the universe expansion rate. Construct a fitness function, calculate the fitness, and determine whether the maximum number of iterations has been reached. If so, obtain the optimized hyperparameters.

15. The method for assessing the condition of an on-load tap changer of a transformer according to claim 14, characterized in that, An adaptive adjustment mechanism is introduced to regulate the parameters of wormhole existence probability E and travel distance rate D. The formula for updating the cosmic expansion rate is as follows: In the formula, Let j represent the j-th parameter of the current optimal universe. , , This represents a random number that follows a uniform distribution and lies between [0, 1]. This represents the probability of a wormhole existing in each generation of the universe, where D represents the travel distance rate. This represents the minimum value of the j-th parameter. This represents the maximum value of the j-th parameter. This represents the j-th parameter of the i-th initialized universe.

16. The method for assessing the condition of an on-load tap changer of a transformer according to claim 14, characterized in that, The fitness function formula is: In the formula, The number of correctly classified samples; This represents the total number of samples.

17. A condition assessment system for an on-load tap changer of a transformer, characterized in that, include: The acquisition unit is used to acquire the electrical performance parameters and mechanical performance parameters of the on-load tap changer of the transformer, wherein the electrical performance parameters include oil chromatography data and the mechanical performance parameters include vibration signals and drive motor current signals; The first generation unit is used to generate an oil chromatographic data feature matrix for the oil chromatographic data, wherein the oil chromatographic data feature matrix includes the volume fraction of gases produced by the decomposition of insulating oil of on-load tap changer under abnormal conditions, total hydrocarbon value, temperature of overheating fault point of on-load tap changer, temperature of discharge fault point, and acetylene growth rate. The second generation unit is used to extract the vibration signal features of the hydropower unit by using multi-scale wave walk entropy for the vibration signal, and to introduce energy entropy, kurtosis and spectral kurtosis to generate a vibration signal feature matrix. The third generation unit is used to generate a drive motor current signal feature matrix for the drive motor current signal, wherein the drive motor current signal feature matrix includes the duration of the drive motor current and the sum of the absolute values ​​of the drive motor current during the motor start-up phase, the motor stable operation phase, and the motor stop operation phase. The fusion unit is used to fuse the oil chromatography data feature matrix, the vibration signal feature matrix, and the drive motor current signal feature matrix to generate a multi-dimensional information feature matrix. The evaluation unit is used to calculate the correlation between data in the multidimensional information feature matrix using an attention mechanism and assign weights to the multidimensional information feature matrix to obtain weighted data samples. The weighted data samples are then input into a pre-built BIGRU model to achieve state evaluation of the on-load tap changer of the transformer.

18. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the transformer on-load tap changer condition assessment method as described in any one of claims 1 to 16.

19. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the transformer on-load tap changer condition assessment method as described in any one of claims 1 to 16.

Images